OCR Paragraph Detection Using Line Classification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Optical Character Recognition (OCR) engines face challenges in processing images with poor quality due to varying contrast, illumination, skew, and distortion, leading to degraded performance and increased processing time, especially when dealing with complex layouts and mixed text/picture content.

Innovation Solution

The OCR process involves paragraph detection, where textual lines are classified into 'beginning' and 'continuation' classes, and pages are categorized as 'poem' or 'text' pages, with alignment detection, using features like line breaks, character width, and indentation to identify paragraphs and determine reading order, employing machine learning techniques for improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional OCR engine processes all pixels in the image, then complete text recognition is achieved, but processing time increases significantly

Engineering Contradiction:
Improvetext recognition accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the image processing task by dividing pixels into text regions and non-text regions. The OCR engine processes only text region pixels, while non-text pixels are skipped. This is achieved through coordinate information that defines bounding boxes around text lines, allowing the system to selectively process only relevant portions of the image, thereby reducing processing time while maintaining recognition accuracy.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If OCR engine processes poor quality images with varying illumination and skew, then adaptability is improved, but performance degrades

Engineering Contradiction:
Improveimage quality toleranceVSAvoidOCR performance
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The patent applies preliminary image preprocessing steps including skew correction, illumination normalization, and contrast enhancement before OCR processing. These preliminary actions prepare the image by correcting geometric distortions and improving visual quality, thereby enhancing OCR performance on poor quality images without requiring the engine to directly handle all image defects during the main recognition process.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If multiple individual processing steps are applied to improve OCR quality, then recognition accuracy is improved, but system complexity increases

Engineering Contradiction:
ImproveOCR qualityVSAvoidprocessing steps
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple processing functions into an integrated system that performs coordinate extraction, text region identification, selective pixel processing, and OCR recognition in a unified workflow. By merging these previously separate steps into a coordinated process, the system maintains high OCR quality while reducing the apparent complexity through streamlined integration of processing stages.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP2545495B1Paragraph recognition in an optical character recognition (OCR) process
Publication Date: 2018.05.30 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP2545495B1 patent drawingFigure 1~3
  • EP2545495B1 patent drawingFigure 4~6
  • EP2545495B1 patent drawingFigure 7~9

AI summary

An image processing apparatus for detecting paragraphs in a textual image includes an input component for receiving an input image in which textual lines and words have been identified and a page classification component for classifying the input image as a first or second page type. The apparatus also includes a paragraph detection component for classifying all textual lines on the input image as a beginning paragraph line or a continuation paragraph line. The apparatus is also provided with a paragraph creation component for creating paragraphs that include textual lines between two successive beginning paragraph lines, including a first of the two successive beginning paragraph lines. The paragraphs that have been identified may be classified by the type of alignment they exhibit. For instance, paragraphs may be classified according to whether they are left aligned, right aligned, center aligned or justified.